Inflated workflow execution graph leads to increase resource consumption and delays

[m4dcoder]

Given the workflow definition at https://github.com/StackStorm/st2ci/blob/86655cb71ae0a1ff813b4bd0828399b54206e146/actions/workflows/st2_pkg_e2e_test.yaml, orquesta converts this workflow into an execution graph for runtime. Due to the number of conditional jumps, multiple references to common tasks, and common tasks having many steps, this leads to an inflated execution graph. This results in very high CPU and memory usage and long delayed when the conductor tries to load the workflow definition. The resource consumption and delay are likely occurring where the conductor is trying to identify start tasks for the graph.